iiRecord
Agentic AI Atlas · Krate Company Brain Memory Query
workflow:krate-memory-querya5c.ai
II.
Workflow JSON

workflow:krate-memory-query

Structured · live

Krate Company Brain Memory Query json

Inspect the normalized record payload exactly as the atlas UI reads it.

File · domain/products/krate.yamlCluster · domain
Record JSON
{
  "id": "workflow:krate-memory-query",
  "_kind": "Workflow",
  "_file": "domain/products/krate.yaml",
  "_cluster": "domain",
  "attributes": {
    "displayName": "Krate Company Brain Memory Query",
    "description": "Workflow for querying and updating the company brain knowledge base\nwithin Krate.\n\nQuery flow:\n1. An agent session (or human user via kubectl) creates an\n   AgentMemoryQuery resource specifying search terms, filters\n   (time range, source repository, author), and similarity threshold.\n2. The aggregated API server resolves the query against the vector\n   index in the AgentMemoryRepository, performing semantic similarity\n   search combined with metadata filtering.\n3. Matching AgentMemorySnapshot resources are returned, ranked by\n   relevance. Each snapshot contains structured knowledge — code\n   patterns, architecture decisions, error resolutions.\n4. The agent incorporates relevant snapshots into its context\n   before proceeding with its task.\n\nUpdate flow:\n1. On agent session completion, the dispatch controller generates\n   a new AgentMemorySnapshot summarizing decisions made, patterns\n   discovered, and lessons learned.\n2. The snapshot is embedded (vector encoding) and stored in the\n   AgentMemoryRepository.\n3. Org-admins can review, curate, or delete snapshots via the\n   memory management UI or kubectl.\n",
    "workflowKind": "development",
    "triggerType": "event-driven",
    "typicalCadence": "per-agent-session",
    "complexity": "moderate"
  },
  "outgoingEdges": [
    {
      "from": "workflow:krate-memory-query",
      "to": "role:platform-engineer",
      "kind": "involves_role"
    },
    {
      "from": "workflow:krate-memory-query",
      "to": "role:backend-engineer",
      "kind": "involves_role"
    },
    {
      "from": "workflow:krate-memory-query",
      "to": "domain:platform-engineering",
      "kind": "applies_to_domain"
    },
    {
      "from": "workflow:krate-memory-query",
      "to": "domain:software-engineering",
      "kind": "applies_to_domain"
    }
  ],
  "incomingEdges": []
}